Abstract
In this article, we argue that because governments fear that accepting mediation and reaching agreements with opponents may signal weakness to other potential challengers, civil war governments tend to resist mediation as a means of demonstrating their resolve. Building on current theories of reputation in civil war, we argue that the threat of future challenges from other groups is likely to be particularly acute in states with multiple ethnic groups, especially states with high levels of ethnic exclusion. We therefore expect that civil wars in these states will be less likely to see mediation and to produce agreements when they do. By examining all instances of mediation in intrastate conflict from 1990 to 2008, we test this argument empirically and find that mediation is more likely to emerge in civil wars where there are a large number of ethnic groups, but is less likely to occur when many of those ethnic groups are excluded from the political process. Once mediation is underway, however, it is less likely to yield a negotiated settlement when there are a large number of ethnic groups, but more likely to end in a negotiated settlement when many of the ethnic groups are excluded from the political process.
Introduction
The civil wars in Syria, South Sudan, and Ukraine underscore the importance (and limitations) of mediation in ending conflicts. Mediation is one of the few tools the international community can deploy to end seemingly intractable civil wars (DeRouen, Bercovitch & Pospieszna, 2011). Unfortunately, the international community often faces barriers to mediating difficult conflicts. As noted by Greig & Regan (2008), while the global community may desperately want some resolution to ongoing civil wars, war-torn governments may be loath to accept mediation. Furthermore, once mediation is accepted, there is no guarantee that a settlement will be reached. Even in cases where the international community has a vested interest in resolving a civil war, mediation may fail to generate a durable solution. As demonstrated by the failure in Syria, mediation efforts may fail despite pressure from regional partners and major powers.
What determines whether international mediation efforts will produce a settlement to ongoing fighting? We suggest that the incumbent government’s concerns for its reputation and the potential for future challenges against it, particularly from politically excluded groups, act as a key barrier to mediation. As noted in Walter’s (2006, 2009a) ground-breaking work on state reputation and the accommodation of secessionist demands, governments are less willing to acquiesce to self-determination movements if they believe that other groups may make similar future claims. Walter suggests that if a government believes it will face future challenges from other groups under its rule, then it is incentivized to develop a reputation for toughness. As a result, governments are less likely to accommodate dissident demands when faced with multiple ethnic groups. Given this finding, what role does government reputation play in the mediation process? While past studies have suggested that reputation plays a crucial role in whether governments accept mediation (Greig & Regan, 2008; Melin & Svensson, 2009), few studies have examined how reputation may influence the success of mediation efforts. If Walter is correct in her assertion that reputational concerns hinder bilateral negotiations, how will they impact the occurrence and success of mediation efforts in civil wars?
We focus our attention on the level of ethnic diversity and exclusion in their societies as reputational concerns of civil war governments. States with large numbers of ethnic groups face more potential fault lines for conflict. To the extent that ethnic groups are politically excluded, this tendency toward violence is even more pronounced. Politically excluded ethnic groups often harbor longstanding grievances (Cederman, Gleditsch & Buhaug, 2013) that make them more likely to see violence as a means to address their grievances. As a result, we suggest that the willingness of civil war governments facing this danger of future violent ethnic group challenges to accept mediation and reach settlements with rebels is conditioned by incentives for governments to foster a reputation for toughness. This reputational concern encourages governments facing an acute threat of future violent challenges from ethnic groups to seek to maintain this reputation for resolve by avoiding mediation, avoiding the signal of weakness that mediation might send to other potential challengers.
These reputational effects also shape the mediation process itself. Even if third parties provide incentives to governments faced with a high potential for future ethnic challengers to overcome their reputational concerns and accept an offer for mediation, there is no guarantee that the government will agree to a negotiated settlement. The government may fear that a peace agreement will signal to other groups in society that the state will make concessions if challenged militarily. In this respect, excluded groups in a state create a pool of actors with significant incentives to adopt violence against their governments, as well as groups with whom governments are less willing to mediate and reach settlements.
We focus our analysis on mediation for two key reasons. First, while mediation, like negotiation, is a voluntary, diplomatic conflict management process, mediation is most likely to be used in long, complex conflicts marked by high levels of animosity between the parties (Bercovitch & Jackson, 2001). Civil wars represent precisely such cases. Second, acceptance of civil war mediation by governments involves bringing in the assistance of a third party, a public signal of willingness to engage in dialogue with a rebel group and consider offering concessions. To the extent that governments are concerned about reputational effects in civil wars, we should be most likely to see this concern reflected in the acceptance and outcomes of civil war mediation.
Our article continues in three parts. First, we provide an overview of previous research examining mediation in civil wars. We note that while the reputation of the government plays a role in previous theories on mediation in civil war, little empirical work has examined how potential reputation costs impact the mediation process. In contrast to previous research, we examine how future reputation costs of the incumbent government shape both the occurrence and success of international mediation. We test these propositions by examining mediation efforts in all civil wars from 1990 to 2005. Interestingly, our empirical findings appear mixed. Mediation is much more likely to emerge in civil wars where there are a large number of ethnic groups, but less likely to occur when many of those ethnic groups are excluded from the political process. On the other hand, once mediation is underway, it is less likely to reach a negotiated settlement when there are a large number of ethnic groups.
Role of mediation in civil wars
The negotiated settlement of civil wars is often hindered by information asymmetries between combatants (Greig & Regan, 2008; Walter, 2009a). Rebels typically foster misinformation during the war to strengthen their bargaining leverage (Mattes & Savun, 2010; Walter, 2009a). Asymmetries of information are especially problematic in civil wars because both sides often have incentives to defect from agreements, making credible commitments to agreements difficult (Walter, 2002). This problem is further compounded by the fact that governments prefer to defeat insurgents as opposed to negotiating with them if there is a chance of victory (Bapat, 2005; Werner & Yuen, 2005). Yet, by refusing to negotiate with rebels, governments also risk that dissidents will gain the ability to win the war. Chad’s President Hissine Habre attempted to show resolve in fighting multiple insurgencies during his rule from 1980 to 1990. Had the government sought a negotiated settlement prior to the proverbial point of no return, it may have been able to ensure its own survival.
Developments during the course of fighting a civil war can open the door for mediation. As combatants spend more time fighting one another, they are able to divine the resolve of their opponents as well as reveal their relative capabilities (Walter, 2009b). As civil wars drag on, combatants become less certain of their ability to win militarily. At the same time, the longer a conflict continues, the more likely a hurting stalemate is to emerge in which the costs and pain of conflict grow increasingly unbearable, encouraging efforts to end the conflict (Zartman, 2000). Given the mistrust among civil war sides as well as the government’s aversion to negotiating with rebels, both sides must typically suffer considerable costs before diplomatic solutions become attractive (Greig & Regan, 2008; Melin & Svensson, 2009; Zartman, 2000). Mediation by a third party can assist civil war belligerents in locating a settlement to their conflict by presenting reliable information to the two sides and helping them overcome commitment problems.
Despite mediation’s benefits, governments face difficult choices when it comes to civil war mediation. Extant research has suggested that accepting mediation can signal weakness to domestic audiences (Beardsley, 2010; Greig & Regan, 2008; Melin & Svensson, 2009). While Beardsley (2010) has suggested that mediators give cover to leaders who may anger their supporters by agreeing to a negotiated settlement, others have argued that the process of bringing in a mediator serves to undermine the authority of the central government. By agreeing to mediation, leaders risk demonstrating a lack of resolve in fighting insurgents and granting legitimacy to armed dissidents. This logic suggests that governments are likely to be acutely aware of the reputational costs inherent to civil war mediation. In the next section, we build on this logic by connecting the reputational costs of civil war mediation to the presence of ethnic divisions within states.
Theory
We explore how reputational costs impact the willingness of governments to accept mediation as well as how these costs shape the likelihood that mediation will settle the conflict. We build on Walter’s (2006, 2009a) argument that governments are less likely to accommodate dissident demands if they fear future similar demands by other groups. Therefore, the greater the number of ethnic groups within a country, the less likely a government is to accommodate the demands of any one dissident group. We argue that if domestic reputation costs shape the willingness of governments to accommodate dissidents, these same reputation costs should also influence the prospects for mediation.
Reputation costs and the occurrence of mediation in civil wars
Walter’s argument addresses a persistent question in the study of conflict: if the costs of fighting a war are high, why are governments unwilling to negotiate with dissidents? Primarily, Walter argues that governments are forward-looking (2006: 313). Elites within the government are not simply concerned with the current demands of dissidents, but also the future demands of other actors in society. Walter assumes that governments are keenly aware that other segments of society suffer asymmetries of information in the bargaining process (Fearon, 1995). Potential dissidents are unsure as to whether the government will accommodate their demands if they engage in contentious politics. Given this incomplete information among dissidents, they predict the future behavior of the government based on past government interactions with other dissidents. By negotiating with dissidents, governments can generate a demonstration effect (Forsberg, 2013; Horowitz, 1985; Lake & Rothchild, 1998), where the success of one group in gaining concessions inspires other groups to make similar demands. Walter (2006, 2009a) argues that governments are aware of this fact. Given the risk that other groups may make similar future demands if the government accommodates aggrieved citizens today, elites are incentivized to avoid negotiating with dissidents. By showing strong resolve, elites attempt to garner a reputation for toughness so as to dissuade future dissent by other groups.
We argue that the government’s desire to signal resolve should also be reflected in its willingness to enter into mediation. While the act of participating in mediation is not the equivalent of offering concessions as part of a settlement, governments do shoulder considerable reputation costs simply by agreeing to engage in mediation. As past research has noted, by entering into formal talks with rebel groups, governments offer legitimacy to armed dissident groups and suggest a lack of resolve (Greig & Regan, 2008; Melin & Svensson, 2009). This legitimacy undercuts the sovereignty of the government, which may embolden other dissidents to engage in armed insurrection so as to garner similar concessions. Furthermore, accepting mediation suggests the government may be willing to offer concessions. As noted by Greig & Regan (2008: 760), ‘[o]ne way to think about the acceptance of mediation is that both sides implicitly agree to make concessions, or at least that they have some concessions to give under certain circumstances’. In other words, agreeing to participate in mediation underscores that governments may be willing to accommodate dissident demands.
If governments fear that other segments of society harbor aspirations to effect political change (either to gain autonomy or greater political power), then these governments should eschew mediation as a means to end fighting. As the acceptance of mediation signals that governments are willing to consider offering concessions as a way to appease armed dissidents, then elites within the government risk signaling a lack of resolve in the face of an armed insurrection. Given that governments are forward-looking, they may prefer to gamble on their success on the battlefield rather than potentially inspiring future rebellion by other groups through engaging in talks during the current conflict. In other words, the costs of sustained fighting may outweigh the potential benefits that mediation offers if governments are concerned that other groups may rebel in the future.
This bloody calculation underscores why Chadian President Hissine Habre preferred to engage in a brutal civil war with multiple rebel factions rather than seek help in ending the conflict. After the 1989 failed coup attempt by Idriss Deby and Mahamat Itno (two of Habre’s commanders), Habre launched a fierce repression campaign against all potential rivals (Atlas & Licklider, 1999). His use of state-directed brutality fell hard on all groups outside of his own Gorane ethnic group. This use of exclusion and repression following a coup attempt is not unusual. The use of fierce brutality helped to dissolve a fragile coalition of other ethnic groups, fostering the emergence of new armed factions throughout the country. Faced with a wide array of dissidents, Habre failed to bring in his international supporters (primarily the US government) to mediate an end to the conflict. Reputation costs may be one key factor that prevented Habre from using international mediation as a tool to resolve the conflict.
Hypothesis 1A: During civil wars, as the number of potential future challengers within a country increases, the willingness of the government to participate in mediation will decline.
We contend that the number of excluded ethnic groups within a civil war state should further provide a barrier to international mediation efforts. Some ethnic groups, all else being equal, are more likely to seek change through armed insurrection than other ethnic groups. It is these specific groups in society that warrant the greatest concern from elites as they consider whether to accept mediation. Drawing on the relative deprivation literature (Horowitz, 1985; Cederman, Wimmer & Min, 2010; Cederman, Gleditsch & Buhaug, 2013), we argue that the number of ethnic groups that are excluded from the political process poses the greatest threat of future insurrection. Given historical discrimination, many of these politically excluded ethnic groups are unable to make claims upon the government through nonviolent institutional mechanisms. Absent conventional methods of seeking concessions from the state, aggrieved ethnic groups should be more amenable to using political violence as a tool to renegotiate the status quo. With this in mind, politically excluded ethnic groups should be interested in whether the government is willing to negotiate when faced with armed dissent. If Walter’s (2006, 2009a) hypothesis is correct, then the government should be especially wary of the number of politically excluded ethnic groups as each aggrieved group may represent a future challenger.
As noted earlier, the process of entering into mediation signals that the government may offer concessions as part of any settlement. We assume that excluded ethnic groups should be particularly interested in signals that the government is willing to accommodate the demands of violent dissidents as excluded groups lack the opportunity to generate political change through conventional avenues. We also assume that the government is aware that excluded ethnic groups are particularly interested in signals that the state may accommodate the demands of rebels. Given these assumptions, we contend that governments should be wary of participating in mediation when they govern excluded ethnic groups.
Hypothesis 1B: During civil wars, as the number of excluded groups within a country increases, the willingness of the government to participate in mediation will decline.
Reputation and the success of international mediation
We also expect that reputation costs will play a role in the success of mediation efforts in generating a settlement to ongoing civil wars. While governments may be uninterested in direct talks with dissidents during conflicts, the international community can place significant pressure on elites to engage in dialogue with rebels. Strong international pressure has forced Syria’s Assad regime to enter into multiple mediated talks with dissidents. Equally, international pressure has often forced Netanyahu’s coalition government in Israel to start talks with Palestinian factions despite the incumbent government’s reservations. We contend, though, that if governments expect that the establishment of a negotiated settlement may inspire other dissidents to engage in armed dissent, elites within the government will fail to offer meaningful concessions, thereby stymieing the mediation process. Third-party leverage may be sufficient to overcome government reputational concerns in joining a mediation effort, but may be insufficient to overcome even deeper regime reputational worries regarding concession-making. This argument assumes that the reputation costs for agreeing to a settlement may be greater than simply entering into mediation. While participating in mediation implies that the government may be willing to offer some concessions, the acceptance of mediation does not guarantee that the concessions that the state offers will be meaningful. Though the government may accept an offer of mediation, it may have no intention of providing real political concessions to mollify dissident concerns.
To overcome this intransigence, mediators have a variety of tools at their disposal. Mediators can facilitate the transfer of credible information between combatants, reducing the role of uncertainty in prolonging civil wars (Kydd, 2003; Svensson, 2009). Mediators can also incentivize civil war settlements by offering inducements (such as economic sanctions or side-payments) (Lake & Rothchild, 1998; Zartman & Touval, 1985). These carrots and sticks can help offset the costs of accommodating the demands of combatants. While rebels often must make some concessions (like demobilizing their forces), the government typically must offer substantial accommodations as part of any settlement. Mediators therefore provide inducements as leverage over the incumbent government to pressure elites to offer meaningful concessions.
This process works only when these inducements outweigh the costs of future rebellion by other groups in society. If the government has little reason to believe that it will be confronted with future challenges as a result of accommodating current insurgents, then the mediator’s inducements should increase the likelihood that elites will reach a settlement with rebels. On the other hand, if the government believes that it will be faced with new armed insurrections as the result of the settlement, then the relative leverage held by the mediator should begin to decline. As the number of potential challengers grows, the ability of mediators to credibly offset the costs of settling the conflict declines, reducing the effectiveness of mediation in ongoing civil wars.
The role of future challengers in shaping the likelihood of mediation success differs from the effects of current challengers against the state. As noted by Nilsson (2010), weaker rebel groups can pressure the government to acquiesce to their demands if there are multiple groups fighting against the state. Rebel groups left out of any settlement that is reached do not necessarily undermine the stability of peace agreements between the government and signatory rebel groups (Nilsson, 2008). Though the number of ongoing conflicts certainly plays a role in shaping the negotiation environment, we contend that the risk of future challengers should hinder the willingness of governments to cooperate with the mediation process. Ongoing conflicts represent known costs to both the mediator(s) and the incumbent government. With that in mind, mediators may adjust inducements to help offset these costs in order to end fighting. On the other hand, the future costs of potential challengers may not be fully understood by participants. Given the uncertainty associated with facing future challengers as a result of any settlement, we expect that governments should be wary of agreeing to any mediated peace agreement.
While the acceptance of mediation implies that governments are willing to offer some concessions (Greig & Regan, 2008), acceptance leaves unknown what types of accommodations the government is willing to offer. By agreeing to a negotiated settlement, however, the government reveals what it is willing to concede to end fighting. The terms of a negotiated settlement reveal the type of demands that the government is willing to accommodate. Given incomplete information, future challengers are unsure as to whether the government will acquiesce to their specific demands if they engage in armed insurrection (Walter, 2006, 2009a). Walter notes that potential dissidents rely on the issue-specificity of accommodations when making the calculation to engage in rebellion. For example, if a government is only willing to offer amnesty or reparations to end an ongoing insurgency, then potential dissidents seeking self-determination will be unsure as to whether they will achieve their objectives if they revolt. Again, Walter suggests that governments are aware that their actions during negotiations will inform the behavior of potential dissidents. By agreeing to specific reforms as part of a peace agreement, governments signal that they may be willing to accommodate similar demands in the future.
Forward-looking governments should therefore be loath to agree to meaningful concessions during mediation if they suspect that they will face similar future challenges. As the number of potential challengers rises, then the establishment of a mediated settlement becomes more difficult. While governments may be forced to accept mediation as a result of international pressure (or as a means to procure more arms), they may perceive the reputation costs as relatively low given that they have not made any genuine or specific concessions. When governments sign on to a peace agreement, they play their proverbial hand, signaling what types of accommodations they would be willing to offer when faced with an armed insurrection.
Hypothesis 2A: During civil wars, as the number of potential future challengers within a country increases, the likelihood of mediation success will decline.
The degree of political exclusion should also play a considerable role in determining whether mediation efforts will be successful in generating a settlement. As noted earlier, despite the government’s concern over signaling weakness to future dissidents, international pressure may force elites to enter into talks with rebels. The start of mediation, though, is no guarantee of success. This is especially true in societies marked by widespread political exclusion. The number of politically excluded ethnic groups presents a more salient threat to elites within the government, as groups who are unable to effect political change through conventional methods are prone to adopt violence (Cederman, Wimmer & Min, 2010). In addition to the threat of new rebellion, excluded groups will also vie for access to the political process. By agreeing to reforms that open up the political process, elites risk alienating their core constituents who may, in turn, remove them from power. In other words, by accommodating the demands of insurgents in societies with broad political exclusion, elites run the risk of facing challenges from outside groups as well as threats within their own ranks. If elites are aware of this risk, then they should seek to stymie the mediation process.
As noted earlier, societies that are marked by widespread political exclusion often ensure the interests of certain segments of society. Elites offer select benefits for their co-ethnics in return for continued political support from these key constituents. These power relations permeate through society, safeguarding the well-being of dominant ethnic groups at the expense of marginalized groups. Disaffected rebel groups should, in turn, seek to reform some of these practices so as to promote public goods that address their grievances. These demands should be reflected in the negotiation process, as rebels seek substantial changes to the political process. By agreeing to these reforms, the government may spark demands from other ethnic groups in society seeking similar concessions (Walter, 2006, 2009a). As the terms of a settlement underscore the types of concessions disaffected groups can expect from the state, the process of agreeing to a settlement (as opposed to merely entering into talks) should have greater salience. Systemic changes that address rebel demands (as well as the demands of potential dissidents) would undoubtedly expand the selectorate size of society, changing the government’s winning coalition (Bueno de Mesquita et al., 2003), and threatening the status of the government’s own co-ethnics. In societies with widespread political exclusion, the perceived costs for accommodating rebel demands through a settlement should be much greater for the government’s constituents.
Beardsley (2010) notes that elites may use the mediation process to provide political cover for unpopular decisions. Increasing the politically relevant strata through accommodations, though, may undermine the efficacy of such political cover, as the government’s core constituents will hold leaders accountable for a loss in relative power. Faced with the prospect of losing their status, the government’s co-ethnics may attempt to spoil the peace process by removing leaders from office to preserve their status in society. Following the emergence of small power-sharing arrangements between the Chadian government and disparate ethnic rebel groups in the late 1980s, Hissine Habre faced a coup attempt from elites within his own ranks (Atlas & Licklider, 1999). Along similar lines, President Nimeiri of Sudan and Prime Minister Rabin both faced assassination attempts following the signing of peace agreements with excluded ethnic groups. Assuming that leaders are forward-looking, then they should be loath to sign any agreement that may expand the size of the selectorate.
Hypothesis 2B: During civil wars, as the number of excluded ethnic groups within a country increases, the likelihood of mediation success will decline.
Research design
Our analysis explores the role that the number of excluded groups in a state plays in both the occurrence of civil war mediation and the outcomes of mediation efforts when they occur. We use data from the Civil Wars Mediation (CWM) dataset (DeRouen, Bercovitch & Pospieszna, 2011) to construct a database with civil conflict months as the unit of analysis. 1 The CWM dataset offers two key benefits for our analysis. First, its operationalization of civil conflict is based upon the definition of civil war used by the UCDP/PRIO data project and it uses conflict beginning and termination dates from the UDCP Conflict Termination dataset (Kreutz, 2010). Second, the CWM dataset provides the most complete coding of both the occurrence and outcomes of civil war mediation efforts currently available to scholars. After excluding cases with missing data, the CWM dataset yields a sample of 6,414 conflict-months distributed across 158 civil conflicts. The temporal domain for our analysis extends from 1990 to 2005. 2
Summary statistics
Our two variables of interest in both the occurrence and outcome of mediation are derived from the Ethnic Power Relations (EPR) dataset (Wimmer, Cederman & Min, 2009). The first variable identifies the number of ethnic groups present in a civil war country. The second theoretical variable identifies the number of politically excluded ethnic groups in each civil war state. We use data from the exclgrps variable in the EPR data as our measure of groups that are excluded from the political process in each country. 5 Table I summarizes the variables in our model.
Although we expect that the number of excluded groups will have an important effect on the occurrence and outcome of mediation in civil wars, we recognize that a number of other factors influence civil war mediation. Because past mediation can encourage subsequent mediation, we use data from the CWM dataset to code a variable describing the number of previous months in which mediation was ongoing in the civil war. Likewise, because the willingness of civil war belligerents to negotiate bilaterally may condition both the need for mediation and the willingness for third-parties to offer mediation, we use negotiation data from Ogutcu-Fu (2016) to create a dichotomous variable identifying those months in which bilateral negotiations were ongoing. Because we are interested in the influence of reputational concerns on the willingness of governments to mediate with rebels, in order to be coded as a 1 a negotiation effort must be conducted sufficiently publicly that the month and year of the talks can be identified.
We use data from the Polity IV dataset (Marshall & Jaggers, 2002) to create a dichotomous democracy variable where states are coded as a 1 during periods of time in which they have a Polity2 score greater than or equal to 6 and a 0 otherwise. We also look to forces linked to the conflict as sources of pressure for mediation. In this respect, we use data from the CWM dataset to create a variable describing the elapsed time in months in which a civil war has been ongoing. We also use data from the UCDP Battle-Related Deaths dataset v5-2015 (Sundberg, 2015) to create a variable describing the cumulative number of battle-deaths produced by the conflict to date. Because more powerful parties tend to feel more capable of winning on the battlefield and imposing terms, making them less willing to mediate, we include measures of both rebel and government military strength in the selection stage of our model. For rebel strength, we use data from the Non-State Actor dataset v3.4 (Cunningham, Gleditsch & Salehyan, 2009) as a measure of the size (in thousands) of the rebel military. Similarly, we use military personnel data from the Correlates of War National Material Capabilities data (Singer, Bremer & Stuckey, 1972) to create a government military personnel variable (in thousands). Because more prosperous states have more resources both to bargain with and buy off rebels, thereby potentially making participation in mediation more attractive, we use data from Gleditsch’s (2002) Expanded Trade and GDP data to include a logged measure of a country’s per capita GDP for inclusion in the selection stage of our analysis. To control for the supply of prospective mediators, we follow the lead of Crescenzi et al. (2011) and include a variable describing the total number of democracies in the international system during each year of our sample. Finally, to account for the non-independence of observations and the binary time-series–cross-sectional nature of our data, we follow the advice of Beck, Katz & Tucker (1998) and include in the selection stage of our model cubic splines and a variable describing the time in months since the last mediation.
In the outcome stage of our analysis, we include several variables describing forces that condition the success of mediation. Because mediation efforts initiated by the parties themselves signal greater buy-in by the belligerents and improve the prospects for successful mediation, we use data from the CWM dataset to include a dichotomous variable in the outcome stage of our model identifying mediation efforts initiated by at least one of the civil war parties. Because democracies may demonstrate different tendencies toward reaching mediated agreement, we include the same dichotomous democracy indicator variable in the outcome stage of our model that is included in the selection stage. We also use CWM data to create a dichotomous variable distinguishing those conflicts that have previously seen multiple mediation efforts from those that have not. Because conflicts over territory may have different prospects for settlement than conflicts over government power, we use NSA data, which bases its issue coding on the UDCP/PRIO Armed Conflict Dataset (Gleditsch et al., 2002), to create a dichotomous variable identifying those conflicts involving a territorial issue. Finally, we include a count variable in the outcome stage of our model describing the number of additional ongoing civil conflicts within a country during a given civil war-month.
Results
Probit model with sample selection
Three cubic splines omitted for space. One-tailed tests. Robust standard errors clustered by conflict in parentheses. **p < 0.01, *p < 0.05, † p < 0.1.
Probability of mediation by number of excluded ethnic groups
Consistent with our expectations, mediation is less likely to occur in civil wars as the number of excluded ethnic groups in the society increases. For an average civil war state with no excluded ethnic groups, the probability of mediation occurring during a civil war-month is approximately .247. 7 As Figure 1 shows, 8 increasing the number of excluded groups to four in that same state drops the probability of mediation during any month to .152, a reduction of nearly 39%. For an average state with eight excluded ethnic groups, the probability of mediation during any given month drops even further to .085, a decline of nearly 66% from the probability of mediation in a similar state with no excluded ethnic groups. In this sense, these results suggest that when civil war governments face more disaffected ethnic groups in their society, the incentives for governments to signal resolve by refusing to bargain with challengers dampens the occurrence of civil war mediation.
Probability of mediation by number of ethnic groups
Interestingly, although the number of politically excluded ethnic groups reduced the likelihood of mediation, contrary to our expectations, the number of ethnic groups in a civil war state increased the chances for mediation. Figure 2 summarizes this effect. For an average state with no politically excluded ethnic groups, there is about a .247 probability of mediation occurring during any ongoing civil war-month. By contrast, a civil war state with 16 ethnic groups, none of which are politically excluded, more than doubles the probability of mediation during a civil war-month, increasing the probability to .563. Increasing ethnic heterogeneity still further to 32 ethnic groups increases the monthly probability of civil war mediation dramatically to .968. These results suggest that ethnic diversity and ethnic exclusion work in two very different ways with respect to the occurrence of mediation. Extensive patterns of ethnic exclusion seem to create a setting in which civil war governments seek to preserve their reputation for toughness by avoiding mediation with challengers to dissuade other politically excluded groups from making similar demands. Multi-ethnic societies with low levels of political exclusion, by contrast, are those in which, by virtue of their inclusion in the political process, an existing pattern of bargaining and compromise over political issues already exists among differing ethnic groups. As a result, this existing history of political inclusion and negotiation acts as a force encouraging governments in ethnically inclusive, heterogeneous states to mediate their civil conflicts.
Figure 3 shows the balancing effects that ethnic diversity and ethnic exclusion have upon the propensity for civil war mediation. At lower levels of political exclusion, the likelihood of mediation grows as the number of ethnic groups in a society increases. For a state with four ethnic groups in which 25% of the groups are politically Probability of mediation by number of ethnic groups and level of political exclusion
With respect to the potential for agreement once mediation takes place, we see some important differences between the effects of the number of politically relevant ethnic groups and the number of excluded groups and the effects we observed in the occurrence of mediation. Probability of agreement by number of ethnic groups (conditional on mediation)
As Figure 4 shows, for an average civil war state with eight politically relevant ethnic groups, none of which are politically excluded, there is a .101 probability that an ongoing mediation effort during the month will yield an agreement. By contrast, the probability of a mediated agreement during a month with mediation drops to .039, ceteris paribus, when a civil war state has 12 politically relevant ethnic groups. At 18 ethnic groups, the probability of an agreement during a month in which mediation is ongoing falls even further to less than .01. Among states with the fewest politically relevant ethnic divisions, the prospects for mediated agreement improve dramatically once mediation takes place. A mediation effort has a .302 probability of success during a given month when directed at a civil war state with two politically relevant ethnic groups and a .396 chance of agreement for a state with no politically relevant ethnic groups.
Evaluating our findings side by side regarding the effects of the number of ethnic groups on the likelihood of mediation and the prospects that mediation will produce an agreement suggests that entry into civil war mediation in the most ethnically pluralistic societies may be seen as a carrying few political costs, unless concessions are ultimately made during these talks. In this respect, for governments in states with many politically relevant ethnic groups, participation in civil war mediation may represent a form of cheap talk, where governments are willing to accept mediation in response to pressure from other states or as a means of improving their domestic and international reputations but with little motivation to produce an agreement as part of the process. A mediated agreement in states with many ethnic groups engaged in the political process represents a far more difficult endeavor than mere participation in mediation, as governments in more pluralistic societies must not only negotiate mediation terms with rebels, but with a broader diversity of political interests than those faced by governments in more homogenous states.
This logic suggests that for more ethnically diverse states, each politically relevant group represents a new veto player in the peace process, hindering the ability of government leaders to craft solutions to contentious issues. Cunningham (2006) observes just such an effect, finding that as the number of veto players involved in civil wars increases, so does the duration of conflicts. Similarly, Thyne (2012) finds that executive constraints placed on leaders during civil wars often hinders the ability of governments to negotiate a settlement to fighting. It makes sense to expect that states with higher levels of ethnic heterogeneity in the political system will see greater restraints, both formal and informal, placed on the executive that will make civil war agreements more challenging.
Although our core interest is in the occurrence and outcome of mediation, our results raise the possibility that if ethnic exclusion reduces the likelihood of mediation, perhaps rather than dissuading governments from talks with rebels, it steers them toward bilateral negotiations. As a robustness check, we reconstructed our analysis to make the occurrence of mediation or negotiation during a civil war-month as the selection stage of our model and the occurrence of a mediated or negotiated agreement as the outcome stage of our model. To do this, we used the same sources of data for mediation and negotiation that we used in the initial model, the CWM data (DeRouen, Bercovitch & Pospieszna, 2011) for mediation and the Ogutcu-Fu (2016) data for negotiation. We use the same 6,414 civil war-month sample used in the preceding analysis. In this sample, we observe 673 civil war-months in which either mediation or negotiation were ongoing. Among these, there were 62 agreements. These results are summarized in Model 2 in Table II. As Model 2 shows, the results when we include negotiation are nearly identical to our findings with respect to the effects of both ethnic diversity and exclusion on the occurrence and outcome of mediation. Our findings may underscore how Walter’s hypothesis applies to all civil war negotiations as opposed to just mediation. While this could be the case, these findings may still inform the international community as to which conflicts are more or less ripe for mediation efforts.
The combination of these results, albeit less than definitive, raises the possibility of a more complicated connection between ethnic exclusion and the outcome of diplomatic efforts. At least for some governments, ethnic exclusion may sometimes make it easier for governments to reach agreements by creating a barrier to entry into talks that, once overcome, actually strengthens the chances for a diplomatic settlement. Under such a scenario, in societies with high levels of ethnic exclusion, once mediation or negotiation is underway, talks create an opportunity for governments to make the costly political accommodations needed to not only end the current conflict but, potentially, prevent subsequent conflicts. Along these lines, Beardsley (2010) highlights the way that government elites can use mediation and pressure from the international community as political cover to make substantial political change without alienating their political base. We believe this suggests a fruitful area for future research.
Putting these pieces together, our findings point to the distinct ways in which the domestic political context of states shapes the way that governments think about reputational effects and how, in turn, this impacts international diplomatic processes to manage civil wars. In many civil war states, governments often bar groups from engaging in the political process. For these excluded groups, this leaves few options outside of violence by which to seek redress of their grievances. Yet, when governments maintain a reputation for an unwillingness to negotiate with rebels, this signals that an excluded group can only achieve its political goals through violence when it is able to achieve outright military victory and impose terms on the government. It is precisely this desire to maintain a reputation for toughness that prevents exclusionary governments from participating in mediation efforts. Governments fear that accepting mediation may signal to other excluded groups that they may achieve their goals by engaging in violence (even if they are unable to win outright on the battlefield). In this sense, in states with high levels of ethnic exclusion, the key barrier to diplomacy is the acute need for governments to develop a reputation for toughness by avoiding talks in the first place.
In ethnically diverse states without ethnic exclusion, the domestic context looks different and shapes the calculus of governments and potential rebels in a different way. For ethnic groups seeking ways to address their grievances with the government, while violence is an option, so, too, is participating in domestic politics. Because domestic politics involves deal-making between governments and domestic groups, the adoption of violence by groups does not result from a lack of access to institutions where government–group bargaining can occur. Instead, the adoption of violence results because groups calculate that they may get better concessions through violence than through politics. Under these circumstances, the core reputational concern for governments in ethnically diverse societies without ethnic exclusion is not an appearance of toughness in accepting talks, but in making concessions. As a result, governments in societies marked by high levels of ethnic diversity (and low political exclusion) should have little problem with accepting international mediation, but will be less willing to offer meaningful concessions. In this way, the key barrier to civil war diplomacy in this environment will not be the initiation of talks, which come with fewer costs for governments, but the reaching of settlements, which potentially bring drastically higher costs for governments.
Conclusion
At the onset of our study we examined what role, if any, the reputation of the incumbent government plays in fostering or hindering mediation in civil wars. We applied Walter’s (2006, 2009a) hypothesis to the occurrence and success of mediation in civil conflicts. As entering into mediation signals that the government is willing to offer some concessions to end the conflict, leaders within the incumbent government may fear that participating in international negotiations will signal a lack of resolve, spurring the onset of future dissident movements. Governments ruling multiple ethnic groups should therefore be loath to enter into mediation during civil wars. Equally, these same reputational concerns should be at play when mediation is underway. We also posit that these reputation concerns should be especially salient when the incumbent government rules over multiple excluded ethnic groups, as excluded ethnic groups are likely to engage in armed dissent given that they lack nonviolent, institutional mechanisms to resolve their grievances.
Our findings suggest that ethnic diversity and ethnic exclusion in civil war states create two very different sets of barriers for mediation. Ethnic exclusion creates a significant barrier to entry into civil war mediation in the first place, but shows no evidence that it dampens the prospects for a mediated agreement once civil war parties can overcome this barrier. By contrast, ethnic diversity, while it serves as a significant barrier to the achievement of mediated agreements, actually increases the likelihood that mediation will occur in the first place. These findings suggest not only that it is the presence of ethnic divisions in civil war states that impacts the mediation process, but also that the way these ethnic divisions manifest themselves within the political system will most significantly influence how they will condition the effectiveness of mediation efforts. This suggests an important avenue for future research to better tease out how the interplay between ethnic divisions and different types of political institutions influences civil war conflict management. While it was beyond the scope of this study, it makes sense to expect that different types of political regimes and different third-party mediation strategies may carry important effects on whether political exclusion serves as a restraining force against government concession-making during mediation or actually encourages it. At the same time, more broadly, our findings highlight the need for future research to examine what role different interests in society play in shaping the occurrence and outcomes of mediation.
Footnotes
Replication data
Acknowledgements
We would like to thank the editorial team at the Journal of Peace Research as well as the anonymous reviewers for their thoughtful questions and excellent suggestions.